In this article self-adjusting spectral clustering method is used to segment Lidar images. In experiments four different Lidar images (sea, forest, desert and city) was used to evaluate the performance of the algorithm. In many cases non-linear without pattern on are intertwined each other and in some cases are inseparable that the evaluation criterion is the ability to distinguish the parts of an optical detection. For each image of the test, eigenvalues of matrix similarity chart is displayed and the optimal number of clusters was estimated. According to the approximate numbers of clusters, similarity matrix patterns a number are divided into groups equal to the number of clusters and clustering was performed. The output of the program, including pictures have label for clusters and each cluster separately from other clusters is displayed. The results showed that, in cases of non-linear region of the cluster together and are intertwined; this method is capable of separating and labeling correctly. This method is a semi-supervised method which is involved in the process of selecting and identifying the user.

Lidar, Three-Dimensional, Normalization, Spectral Clustering, Self-Regulated